F. Wessels wrote:
> I do agree that WOL is not provided by the OS. The OS isn't running after
> all. But the OS does, as Randy mentioned, needs to provide an interface to
> control the phy to leave it in a state where it can respond to a magic
> packet. I've observed this on the same hardware but different OSes. I did
> test it with the e1000g driver on both pci and pcie but it turns off the phy.
> If I should report this in a different group than networking please let me
> know.
>
Its more than just programming the PHY. You need to program the address
filter to decide *which* conditions (what kinds of packets) you want to
cause the machine to wake up. (And, optionally, do you want link state
change to also wake the machine up?)
There are more than one form of "magic" packet involved, and in some
cases NICs also offer features like security checks in magic packets.
The bigger question then becomes *how* do we offer the capability to
configure WOL to administrative users. *Then*, once a NIC driver knows
its doing WOL, it can do the right thing and not shutdown the PHY.
(Right now the NIC drivers just assume that the system is losing power,
and disable not just interrupts, but the PHY, mac controller, etc.)
All this needs to be the product of some future project. Its not just a
couple lines of change in a driver somewhere.
-- Garrett
>
_______________________________________________
networking-discuss mailing list
[email protected]
